A young physician named Ignaz Semmelweis set out to find the cause.

His first guess was posture. Women in the midwife ward tended to give birth on their side, women in the doctor ward on their back. He asked the doctors to change position. The death rate did not move.

His second guess was fear. A priest walked through the doctor run ward ringing a bell whenever a woman died, and Semmelweis wondered whether the sound itself was killing patients through sheer dread. He asked the priest to take a different route and stop ringing the bell. Again, there was no improvement.

Both guesses were visible, testable, close enough at hand to try. It took a death outside the ward entirely to break the pattern. A colleague of his, a professor named Jakob Kolletschka, died after an accidental cut on his finger during an autopsy. Kolletschka’s own autopsy showed the same signs Semmelweis had seen in the women who died of childbed fever. Semmelweis realized the doctors in his ward were coming directly from performing autopsies to delivering babies, without washing anything off their hands in between. He did not have a word for what was being carried over but he had a mechanism. He ordered every doctor to disinfect their hands with chlorinated lime before examining a patient, every time, no exceptions. Within a year, the mortality rate in the doctor run ward fell to 1.27 percent, lower than the midwife run ward’s own rate of 1.33 percent that same year.

Before going any further, it helps to be direct about why a story from a Vienna maternity ward in the 1840s should matter to someone who has never worked in a hospital and never will. Most people are carrying a version of posture or a version of the bell right now, just in something closer to home. An explanation gets picked for why a relationship keeps hitting the same wall, why a habit will not stick, why a project at work keeps stalling in the same place, and it is rarely picked because it was tested and held up. It is picked because it was the first thing close enough to reach for, and it survives mostly because nobody has acted on it directly enough for it to fail out loud. Try harder at the habit and it still slips by the second week. Fix the one thing in the relationship that seemed obvious and the same fight comes back a month later anyway. Nothing about the explanation was disproven before that. It just had not been tested yet, the way posture and the bell had not been tested until someone changed them and the death rate refused to move. What follows is a way of catching that kind of guess earlier, before it costs what Semmelweis’s guesses cost him.

A separate practice of mine is keeping twelve questions open on purpose rather than settled, ones that stay interesting long after any answer arrives, described in Twelve Questions, Invited in for Tea. Mental models turn out to do two different jobs for a question like that, tools for searching out a mechanism, and a test for whether a question can accept one at all, and Semmelweis’s ward is where both jobs are easiest to see.

What a Mental Model Is

A mental model is an internal representation of how something behaves, built by watching a mechanism operate closely enough to reason with it and predict what happens next, even in situations never directly observed. It does not tell anyone what to do. Knowing that something on unwashed hands could travel from a corpse to a patient is a mental model. It does not say when to wash, or how often, or in what solution. It only says how the transfer happens, in any ward, on any morning.

1. Mental Models as Reasoning Tools

The version most people use day to day is Shane Parrish’s cataloguing of recurring mental models: general reasoning moves like inversion, first principles, and Occam’s razor, ways of thinking through any problem, not pictures of how one specific thing works. None of the three originate with Parrish, but his cataloguing is where most people meet them, and they are the tools that got Semmelweis to a candidate mechanism in the first place. Semmelweis’s ward is a good place to see what each one does.

Inversion. The natural question was what is killing the doctors’ patients. The more useful question, and the one Semmelweis was slow to ask, was why do the midwives’ patients keep living. Posture and the bell were both still answers to the first question. Turning the question around did not hand him the mechanism directly, but it pointed him at the right comparison, the one ward that was somehow not a problem, rather than only interrogating the one that was.

First principles. The inherited convention of the time was miasma theory, the idea that bad air in general caused disease. Reasoning from that convention would have kept Semmelweis looking at ventilation, crowding, the general atmosphere of the ward. Reasoning from what was directly observed, one ward killing at a much higher rate than the other with nothing else obviously different between them, is what put the miasma explanation aside entirely rather than trying to patch it.

Occam’s razor. Posture and the bell are often misremembered as the simple explanations, discarded in favor of something more complicated. They were not simple, they were just close at hand. The simplest explanation consistent with everything observed was the one difference nobody had thought to look at, what the doctors were doing with their hands in the hour before they walked in.

The same three moves generalize past this one hospital. When an explanation for a stalled habit or a recurring argument feels obvious the moment it is reached for, invert the question first: instead of asking what is causing the problem, ask what is different about the one version of the situation where it does not happen. Reason from what can be observed rather than from whatever explanation happened to be inherited, the way miasma theory was inherited before Semmelweis set it aside. And when two explanations both fit, take the one that requires checking something nobody has looked at yet, not the one that happens to be sitting closest at hand.

2. Mental Models as a Sorting Test

The second is different: a mental model can also test whether a question is the kind that has a mechanism to find in the first place, which is where the twelve favorite questions come back in.

The mechanism, once verified, opened onto a different set of questions entirely, the kind that did not close when the number did. Putting myself in his shoes, here is what five of my own twelve favorite questions, using the numbering from the original list, would have asked next.

Question 7. How to shape an idea so it is ready to be received. He had the data. He published it, sent copies to prominent physicians across Europe, and watched most of them dismiss it anyway. Having the answer, and in retrospect the right answer, did not mean he had found a way to make anyone else hold it.

Question 5. Given any problem, how to find the question worth answering. The diagnostic question, what is killing these women, he answered in months. The question he never seemed to resolve was different: whether the fight worth having was convincing Vienna’s medical establishment, or simply running his own ward correctly and letting the results speak for themselves. He kept choosing the fight.

Question 4. What is the objective function. He said, more than once, that he wanted no personal credit, only the practical outcome of a mother keeping her child. But the letters he wrote later are furious and accusatory in a way that reads like a man chasing something else too, being heard, being vindicated, maybe just being believed by someone. Those are not the same objective, and it is not obvious he ever picked one.

Question 12. Who are his people. Johann Klein, his own superior, looked at a falling mortality rate and credited a new ventilation system, on a ward whose ventilation had not changed at all, taking the thing that happened to change around the same time and calling it the cause without asking what else moved. Klein held real power over Semmelweis’s career and used it. When Semmelweis’s assistant professorship came up for renewal in 1849, Klein did not renew it. Whoever Klein was to him, he was not one of Semmelweis’s people, and losing him meant losing the answer to who counted as his people at all, once the men who trained him had decided he was the problem rather than the finding.

Question 11. What does he owe the world. Physicians across Europe kept dismissing the idea, year after year, even with the book in hand. He knew something that could save lives, and the people with the power to act on it would not. None of it resolved. His mind gave out under the weight of carrying it, and he died in an asylum in 1865, accepted only once he was no longer there to know it.

Not all five of these hold the same kind of resistance, and separating which do from which do not matters here. Some have an actual mechanism to grip, something with parts that push against each other, not sitting side by side, something that still explains a case it was never built for. Others do not, no matter how carefully the description was built.

Questions four and eleven, what is the objective function and what does he owe the world, are values questions: they have nothing for a mechanism to grip in the first place. There is no candidate explanation to test, because there is nothing describing how something works, only a claim about what ought to be wanted or what is owed. A perfectly built model of how credit and recognition operate inside a medical hierarchy would still not tell Semmelweis whether he should have wanted vindication or only outcomes. That is a decision, not a description, and no mechanism, however well tested, was ever going to settle it.

Questions seven and twelve are a different case: call them mechanism questions. This is where a real explanation holds up. Take who are his people. A candidate mechanism is available: people defend a position when the evidence backing it outweighs the cost to their own standing, and abandon it when the cost outweighs the evidence. That is two things pushing against each other, evidence on one side, reputational cost on the other, and it does not only describe Klein, it describes any colleague deciding whether to defend or abandon a position once their own name is attached to being wrong. It does not decide who Semmelweis’s people were. It describes how people in his position tend to behave once their own standing is on the line, which is close enough to an answer that a person could act on directly. Shape an idea so it is ready to be received holds up the same way: an idea’s chance of adoption comes down to how much it threatens existing authority weighed against how credible its messenger already is, and that holds well past this one hospital. Question five, given any problem how to find the one worth answering, sits in between, part describable pattern, part a matter of worth that no description ever settles.

A guess and a mechanism can feel identical from the inside, right up until one of them gets tested and the other does not. Posture and fear felt exactly as reasonable to Semmelweis, in the moment, as cadaverous particles did once he found them. Some of my own twelve behave like seven and twelve, mechanism questions informed by something real nearly to the point of answered. Others behave like four and eleven, values questions that no mechanism will ever close, because they were never asking how something works to begin with. What a real mechanism does for a question in that second group is not close it. It gives the question something solid to push against instead of instinct and hurt feelings alone, the same way a confirmed mechanism let the question about Klein finally get asked in specific, checkable terms rather than staying a vague grievance. A guess wearing a model’s clothes cannot do that. It only feels like it can, right up until it is tested. That is the difference that matters, not whether a favorite question ever gets answered, since some of them never will, but whether the ones that stay open get engaged with something real.

None of this needs a hospital, or a hundred and eighty years of hindsight, to put to use. It needs being honest about which explanation currently running a person’s own decisions is a posture guess and which is an actual mechanism, and being willing to find out which is which before the cost of being wrong grows the way Semmelweis’s did.

A Closing Reflection. Posture and the bell felt just as reasonable as the real cause, right up until someone tested them. The four guesses below are a chance to run that same test on something smaller and much more ordinary.

1. Where do you guess instead of measure, and get away with it most of the time: timing how long something will take, how much to pack, how much to say in a message before it reads as too much or too little, how long to wait before following up on something.
2. Where has someone else corrected your approximation, and the correction is what stayed rather than the original guess: a moment someone said no, it is not like that, here is why, and the reason stuck with you afterward.
3. Where do you approximate on purpose, knowing you are doing it, because the real thing does not earn precision, versus where you approximate without noticing and only find out later that it cost you something.
4. Where do you approximate specifically because you do not care to do otherwise, not from ignorance but from a deliberate choice not to invest attention there, a slightly more honest version of eyeballing than the usual story of someone who wanted precision and did not know how to get it.

Sit with those for a minute rather than answering right away. The one that gives a specific afternoon, a specific voice, a specific object in hand is the one worth zooming into. Sorting out which one it is isn’t the mental model itself. It is the first step toward one, the same way noticing that the two wards differed only in what the doctors did with their hands was a first step, not yet the germ theory that followed. Search for the mechanism the way inversion, first principles, and Occam’s razor did. A mechanism, once recognized as one, is worth continuing to press on.

Where This Practice Came From

Craik proposed in 1943 that the mind builds small working models of reality to test outcomes before acting, and Johnson-Laird’s later work kept the core of it: a mental model is a representation used to reason and predict, not just a summary, one whose parts push against each other and that still holds up in a case it was not built for. Parrish’s cataloguing of recurring mental models, drawn together across his Farnam Street work, is where most people encounter tools like inversion, first principles, and Occam’s razor today, though none of the three originate with him. Senge’s The Fifth Discipline named mental models as one of the disciplines of a learning organization, the often unexamined images of how the world works that quietly shape action.

Intellectual Honesty Note. The mechanism question and values question naming, and the framing of mental models as doing two separate jobs, a reasoning tool and a sorting test, is my own construction, not inherited directly from Craik, Johnson-Laird, Parrish, or Senge.

References

Craik, K. J. W. (1943). The Nature of Explanation. Cambridge University Press.

Gombos, Róza. “Ignaz Semmelweis.” Science History Institute, September 15, 2025. https://www.sciencehistory.org/education/scientific-biographies/ignaz-semmelweis/

Johnson-Laird, P. N. (1983). Mental Models: Towards a Cognitive Science of Language, Inference, and Consciousness (No. 6). Harvard University Press.

Johnson-Laird, P. N., Girotto, V., & Legrenzi, P. (1998). Mental Models: a Gentle Guide for Outsiders. Sistemi intelligenti, 9(68), 33.

Johnston, Katrina. “Ignaz Semmelweis: The Saviour of Mothers.” Canadian Museum of Health Care, 2022. https://www.museumofhealthcare.ca/blog/ignaz-semmelweis-the-saviour-of-mothers

Parrish, S. (2019). The Great Mental Models: General Thinking Concepts. Latticework Publishing.

Senge, P. M. (2006). The Fifth Discipline: The Art and Practice of the Learning Organization. Broadway Business.